INDEX

Explanations

It is challenging to determine a specific pattern as neuron 4 seems to be activated by a series of special characters and letters without a clear relation

oai_token-act-pair · gpt-3.5-turbo

special characters and symbols, possibly indicating formatting or encoding anomalies in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Constantine

-0.86

 ponies

-0.79

 Harper

-0.79

 Arkham

-0.78

 settlements

-0.76

 Cantor

-0.76

 Bethlehem

-0.76

 landowners

-0.75

 Alexandria

-0.73

 directories

-0.73

POSITIVE LOGITS

ë

2.36

ì

2.34

í

2.28

ê

2.27

ëĭ

2.26

ìĿ

2.23

íķ

2.22

ì

2.13

ë

2.12

é¾įå

1.43

Activations Density 0.008%

It is challenging to determine a specific pattern as neuron 4 seems to be activated by a series of special characters and letters without a clear relation

special characters and symbols, possibly indicating formatting or encoding anomalies in the text

No Comments

No Known Activations

It is challenging to determine a specific pattern as neuron 4 seems to be activated by a series of special characters and letters without a clear relation

special characters and symbols, possibly indicating formatting or encoding anomalies in the text

No Comments

No Known Activations